Running AI Pentests in Production: Reflectiz Founders Explain

When you let an AI agent hack your production website, what happens if something breaks? Reflectiz co-founders Ysrael Gurt and Idan Cohen talk through how they built an agentic penetration testing agent designed to run safely against live production environments.

What's covered:

• Why the "brain" behind an AI agent matters far less than the harness built around it
• Why Reflectiz built a model-agnostic pentesting agent instead of betting on a single LLM
• What it actually takes to run an agentic pentest safely against a live production website
• Why full attack coverage matters more than any single vulnerability found
• How layered guardrails stop an agent from ever sending a harmful payload
• Why scan depth and pricing need to be adjustable, from a quick daily check to a full "expert mode" scan
• A real example of the agent finding a critical authentication bypass on a production finance site

Speakers: Ysrael Gurt, CTO & Co-Founder, Reflectiz; Idan Cohen, CEO & Co-Founder, Reflectiz

Key terms: agentic pentesting, continuous penetration testing, AI security agent, production penetration testing, offensive security, AppSec, PCI DSS

See Offensive Hub in action at https://www.reflectiz.com/blog/agentic-pentesting-offensive-hub-walkthrough/. Worth a listen if you're weighing whether AI pentesting agents are ready to run against your live site.